Please use this identifier to cite or link to this item: http://hdl.handle.net/10419/79261
Authors: 
Ramalho, Esmerelda A.
Smith, Richard J.
Year of Publication: 
2003
Series/Report no.: 
cemmap working paper, Centre for Microdata Methods and Practice CWP07/03
Abstract: 
Missing values are endemic in the data sets available to econometricians. This paper suggests a unified likelihood-based approach to deal with several nonignorable missing data problems for discrete choice models. Our concern is when either the dependent variable is unobserved or situations when both dependent variable and covariates are missing for some sampling units. These cases are also considered when a supplementary random sample of observations on all covariates is available. A unified treatment of these various sampling structures is presented using a formulation of the nonresponse problems as a modification of choice-based sampling. Extensions appropriate for nonresponse are detailed of Imbens' (1992) effcient generalized method of moments (GMM) estimator for choice-based samples. Simulation evidence reveals very promising results for the various GMM estimators proposed in this paper.
Subjects: 
Generalized Method of Moments Estimation , Missing Completely at Random , Nonignorable Nonresponse , Semiparametric Efficiency
JEL: 
C25
C51
Persistent Identifier of the first edition: 
Document Type: 
Working Paper

Files in This Item:
File
Size
596.28 kB





Items in EconStor are protected by copyright, with all rights reserved, unless otherwise indicated.